KR20220002679A - Machining condition search device and machining condition search method - Google Patents

Abstract

The processing condition search apparatus 2 according to the present invention includes a processing condition generating unit 22 that generates processing conditions composed of one or more control parameters settable in the processing machine 1, and processing based on the generated processing conditions. A thread processing command unit 23 to cause the processing machine 1 to perform; a processing evaluation unit 12 for generating an evaluation value of the processing performed based on information indicating a processing result of the processing performed; , based on the processing condition corresponding to the evaluation value, the prediction unit 27 for predicting the evaluation value corresponding to the processing condition in which processing is not performed, the predicted value predicted by the prediction unit 27, and the processing evaluation unit An optimum machining condition calculation unit 31 is provided for obtaining, based on the evaluation value generated by (12), an optimum machining condition, which is a machining condition in which the evaluation value is equal to or greater than the threshold value and the allowable error is maximized.

Description

Machining condition search device and machining condition search method

The present invention relates to a machining condition search device and a machining condition search method for searching for machining conditions set in a machining machine to obtain an appropriate machining result.

A processing machine used for industrial purposes performs processing to change the shape of a workpiece that is a workpiece. The material of the workpiece is various, such as metal, ceramic, glass, and wood. An example of the processing is a removal processing in which a workpiece is formed into a desired shape by removing an unnecessary portion. In the removal process, unnecessary parts are removed, for example, by procedures, grinding, or the like, or by using electricity or other energy.

In general, a plurality of control parameters can be set for these processing machines. The processing result of the processing machine depends on processing conditions that are combinations of respective parameter values of a plurality of control parameters. That is, in order to obtain a desired processing result, it is necessary to set processing conditions suitable for a processing machine. However, since there are a plurality of control parameters and the parameter value of each control parameter can be set as a continuous value or in multiple steps, it takes a lot of effort and time to select a processing condition such as to obtain a desired processing result by trial. in need.

For example, in a metal plate laser processing machine that uses laser light to cut a work piece called mild steel of about 5 to 20 mm, even if there is a variation in the size of the work piece, the cut surface formed by machining is of the required quality. that is being demanded In the case of a metal plate laser processing machine, there are five main control parameters that have a large influence on the processing result: laser power, cutting speed, beam magnification, focus position, and gas pressure. Each control parameter is selected from among values of a plurality of steps. Here, assuming that each of the five control parameters can be selected from the values of 11 steps, the total number of combinations is 16,1051. At this time, if 5 minutes are required to execute one processing condition, about 559 days are required for the execution of 161051 types.

In this way, in order to select appropriate processing conditions, it takes time and effort to perform trials in all combinations, so generally, trials of all combinations are not performed, and appropriate processing conditions are selected by another method. As an example of another method, the method of patent document 1 is mentioned. In the method described in Patent Document 1, a learning model showing the relationship between processing conditions and evaluation values is generated by a functional model such as a neural network, and the evaluation value is the highest using the prediction result by the learning model. Explore the processing conditions to be used. Thereby, in the method of patent document 1, processing conditions with a high evaluation value can be obtained with a small number of trials.

[Patent Document 1] Japanese Patent Laid-Open No. 2008-036812

In the method of patent document 1, based on the combination of the processing conditions and evaluation value obtained by trial, the optimal processing conditions are calculated|required by searching the periphery of the processing conditions with the highest evaluation value. However, in general, even if the set machining conditions are the same, the machining results may differ depending on factors other than the machining conditions. For example, even if the same processing conditions are set due to variations in the size of the material to be processed, variations may occur in the processing results. For this reason, even if it sets the processing conditions used as the high evaluation value selected by the method of patent document 1 to a processing machine, if processing is actually performed, the processing result more than a desired evaluation value may not be obtained.

The present invention has been made in view of the above, and an object of the present invention is to obtain a processing condition search device capable of obtaining a processing condition with a high possibility of obtaining a processing result equal to or greater than a desired evaluation value.

In order to solve the above problems and achieve the object, a processing condition search apparatus according to the present invention includes a processing condition generation unit for generating processing conditions composed of one or more control parameters settable on a processing machine, and a processing condition generation unit and a thread processing command unit for causing the processing machine to perform processing based on the processing conditions generated by the . In addition, the processing condition search device includes a processing evaluation unit that generates an evaluation value of the processing performed based on information indicating a processing result of the processing performed, and a processing evaluation unit that generates an evaluation value of the processing performed, based on the evaluation value and a processing condition corresponding to the evaluation value, A prediction unit for predicting an evaluation value corresponding to a processing condition that has not been implemented is provided. In addition, the processing condition search apparatus determines, based on the prediction value predicted by the prediction unit and the evaluation value generated by the processing evaluation unit, the optimum processing condition, which is a processing condition in which the evaluation value is equal to or greater than the threshold value and the allowable error is maximized. An optimum processing condition calculation unit to be obtained is provided, and the evaluation value indicates that the larger the value, the better the processing result.

The processing condition search apparatus according to the present invention is effective in that processing conditions with a high probability of obtaining a processing result equal to or greater than a desired evaluation value can be obtained.

1 is a diagram showing a configuration example of a processing condition search apparatus according to an embodiment of the present invention;
2 is a diagram showing a configuration example of a processing circuit according to an embodiment;
3 is a flowchart showing an example of a processing condition search processing procedure in the processing condition search apparatus according to the embodiment;
4 is an explanatory diagram conceptually indicating an index indicating a predicted value of an evaluation value of the embodiment and an uncertainty;
5 is a view for explaining the tolerance of the embodiment;
It is a figure which shows the display example of processing conditions and evaluation value of embodiment.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a machining condition search device and a machining condition search method according to an embodiment of the present invention will be described in detail with reference to the drawings. In addition, this invention is not limited by this embodiment.

(Embodiment)

BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows the structural example of the processing condition search apparatus which concerns on embodiment of this invention. The processing condition search apparatus 2 of the present embodiment searches for processing conditions suitable for processing of the processing machine 1 . As shown in FIG. 1 , the processing condition search apparatus 2 of the present embodiment includes a search processing condition generation unit 11 , a processing evaluation unit 12 , a machine learning unit 13 , and an optimal processing condition generation unit ( 14) and a display unit 15 .

The processing machine 1 is, for example, an industrial machine which completes a workpiece|work which is a to-be-processed object into a desired shape by removing an unnecessary part, or an industrial machine which performs additional processing, etc. FIG. Hereinafter, the work is referred to as a work. An example of the material of the work is metal, but the material of the work is not limited to metal and may be ceramic, glass, wood, or the like. As such a processing machine 1, a laser processing machine, an electric discharge processing machine, a cutting processing machine, a grinding processing machine, an electrolytic processing machine, an ultrasonic processing machine, an electron beam processing machine, an additional processing machine, etc. correspond. In this embodiment, although the processing machine 1 is demonstrated as a laser processing machine, this invention is applicable also to processing machines other than a laser processing machine.

When the processing machine 1 is a laser processing machine, the processing machine 1 cut|disconnects a workpiece|work as follows, for example. The processing machine 1 which is a laser processing machine is equipped with a laser oscillator and a processing head. The laser oscillator oscillates and emits laser light. The laser beam emitted from the laser oscillator is supplied to the processing head through an optical path. A processing gas is supplied to the inside of the processing head, and when a laser beam is irradiated to the workpiece, the processing gas is supplied to the workpiece. The processing head has a condensing lens for condensing laser light onto a workpiece. The processing head cuts the workpiece by condensing the laser beam and irradiating the workpiece. In this way, the workpiece is cut by the processing machine 1 .

The processing machine 1 can perform normal processing for making a workpiece|work into a desired shape, and can perform the experimental processing mentioned later on a workpiece|work. In the processing for experiments, the processing machine 1 of the present embodiment applies preset processing for experiments to the workpiece according to the processing conditions for search, which will be described later, generated by the search processing condition generation unit 11 . The processing machine 1 is equipped with the sensor etc. which can acquire the information which shows a processing result. The information indicating the processing result is, for example, information indicating the state of the processing machine 1 or the workpiece during processing, and information indicating the state of the workpiece after processing. Such information is acquired by the sensor etc. with which the processing machine 1 is equipped. An example of the sensor is a sensor that detects a sound, light, or the like generated during processing. Another example of the sensor is an imaging device that acquires an image obtained by imaging the workpiece after processing, and a measuring instrument that measures the unevenness of the cut surface of the workpiece. The processing machine 1 should just be equipped with the sensor which can detect the information which shows a processing result, and a sensor is not limited to these. In addition, although the example in which the processing machine 1 is equipped with a sensor is demonstrated here, a sensor may be provided separately from the processing machine 1, and you may make it the processing condition search apparatus 2 acquire a detection result from this sensor.

The search processing condition generation unit 11 of the processing condition search device 2 generates processing conditions used in real processing for experiments, and causes the processing machine 1 to perform processing based on the generated processing conditions. That is, the search processing condition generation unit 11 generates the processing conditions searched for by actual processing in a multidimensional space in which the control parameters constituting the processing conditions are dimensioned. The processing conditions include a plurality of control parameters that can be adjusted with respect to the processing performed by the processing machine 1 . Control parameters are, for example, laser power, cutting speed, beam magnification, focal position, and gas pressure. As shown in FIG. 1 , the search processing condition generation unit 11 includes a search end determination unit 21 , a processing condition generation unit 22 , and a real processing command unit 23 .

The machining condition generating unit 22 generates machining conditions composed of one or more control parameters that can be set in the machining machine. In detail, the processing condition generation unit 22 generates processing conditions used in processing for experiments, and outputs the generated processing conditions to the actual processing command unit 23 . The thread processing command unit 23 causes the processing machine 1 to perform processing based on the processing conditions generated by the processing condition generation unit 22 . In detail, the real processing command unit 23 generates a command for operating the processing machine 1 based on the processing conditions input from the processing condition generation unit 22 , and transmits the generated command to the processing machine 1 . output to The search end determination unit 21 determines whether or not to end the experimental processing based on the information used for the end determination, and outputs the determination result to the processing condition generation unit 22 . The information used for the end determination is generated by the machine learning unit 13 as will be described later. Further, in the present embodiment, the machining condition search device 2 has a function as a control device for outputting a command to the machining machine 1 , and generates a machining command based on the machining condition and outputs it to the machining machine 1 , have. It is not limited to this example, You may provide the processing condition search apparatus 2 separately from the control apparatus which outputs the instruction|command to the processing machine 1 . When the machining condition search device 2 is provided separately from the control device that outputs a command to the machining machine 1 , the machining condition search device 2 does not need to include the actual machining command unit 23 . In this case, the processing condition search device 2 is connected to the control device described above, and the search processing condition generation unit 11 outputs the generated processing conditions to the control device described above.

The processing evaluation unit 12 generates an evaluation value of the performed processing based on the information indicating the processing result of the performed processing. In detail, the processing evaluation unit 12 acquires information indicating the processing result from the processing machine 1 , obtains an evaluation value indicating the quality of processing using the acquired information, and evaluates processing conditions and corresponding It remembers the search result, which is a combination of values. The processing evaluation unit 12 includes a processing result acquisition unit 24 , an evaluation value generation unit 25 , and a search result storage unit 26 .

The processing result acquisition unit 24 acquires information indicating the processing result from the processing machine 1 . Based on the information acquired by the processing result acquisition part 24, the evaluation value generation|generation part 25 calculates|requires the evaluation value with respect to the processing performed based on the processing condition for each processing condition. An evaluation value is a value from 0 to 1, for example, and it defines as a value which shows that a processing result is so good that a value is large. Therefore, when the best machining is performed, the evaluation value becomes 1, and when the worst machining is performed, the evaluation value becomes 0. Good processing varies depending on the type of the processing machine 1, the specifications required for the workpiece to be processed by the processing machine 1, etc. For example, when cutting with a laser processing machine is performed, the cut surface may be scratched or rough. It indicates that there is little or no processing defects such as load, oxide film peeling, and dross. Dross is a phenomenon in which molten metal or the like adheres to the lower surface of the workpiece. Roughness shows that the surface precision of a cut surface is bad. The evaluation value may be calculated based on the information acquired by the processing result acquisition unit 24 by the evaluation value generation unit 25, or may be determined based on input by a user. In the latter case, for example, the evaluation value generation unit 25 presents the information acquired by the processing result acquisition unit 24 to the user by displaying it on a display unit (not shown) or the like, and receives the information from an input unit (not shown). , to obtain the evaluation value input by the user using the input means. The details of the operation of the evaluation value generating unit 25 will be described later.

In addition, the evaluation value generating unit 25 stores the combination of the processing condition and the evaluation value in the search result storage unit 26 as a search result. The search result storage unit 26 stores the search result.

The machine learning unit 13 uses the search result stored in the search result storage unit 26 to predict the evaluation value of machining corresponding to the unsearched machining condition. Moreover, the machine learning part 13 calculates the uncertainty with respect to the predicted value of an evaluation value, ie, prediction easiness.

The machine learning unit 13 includes a prediction unit 27 , a prediction result storage unit 28 , an uncertainty evaluation unit 29 , and an uncertainty storage unit 30 . The prediction unit 27 predicts the evaluation value corresponding to the processing condition to which processing is not performed, based on the evaluation value generated by the evaluation value generation unit 25 and the processing condition corresponding to the evaluation value. Specifically, the prediction unit 27 uses the search result stored in the search result storage unit 26 to predict the machining evaluation value corresponding to the unsearched machining condition, that is, the unsearched machining condition, , the predicted values of the evaluation values obtained by prediction are stored in the prediction result storage unit 28 in association with the processing conditions. The prediction result storage unit 28 stores, as a prediction result, a combination of the unsearched processing condition and the predicted value of the corresponding evaluation value. The uncertainty evaluation unit 29 calculates an index indicating the uncertainty of prediction by the prediction unit 27 . In detail, the uncertainty evaluation unit 29 uses the search result stored in the search result storage unit 26 to calculate the uncertainty about the predicted value of the evaluation value, that is, an index indicating that the prediction is easy to deviate, The calculated index is matched with the processing conditions and stored in the uncertainty storage unit 30 as uncertainty information. The details of the operations of the prediction unit 27 and the uncertainty evaluation unit 29 will be described later.

The optimum processing condition generation unit 14 determines that the evaluation value is equal to or greater than a threshold value and the allowable error is the maximum based on the prediction value predicted by the prediction unit 27 and the evaluation value generated by the processing evaluation unit 12 . Find the optimal processing condition, which is the processing condition for The allowable error will be described later. In detail, the optimum processing condition generation unit 14 performs processing of the processing machine 1 based on the prediction result stored in the prediction result storage unit 28 and the uncertainty information stored in the uncertainty storage unit 30 . It creates an optimal processing condition that is an appropriate condition for

The optimum machining condition generation unit 14 includes an optimum machining condition calculation unit 31 and an optimum machining condition storage unit 32 . The optimum processing condition calculation unit 31 is configured to determine a condition suitable for processing of the processing machine 1 based on the prediction result stored in the prediction result storage unit 28 and the uncertainty information stored in the uncertainty storage unit 30 . An optimum machining condition is generated, and the generated optimum machining condition is stored in the optimum machining condition storage unit 32 . The details of the operation of the optimum processing condition calculation unit 31 will be described later. The optimum machining condition storage unit 32 stores optimum machining conditions.

The display unit 15 reads and displays the information stored in each storage unit according to a request from the user. The display unit 15 displays, for example, the evaluation value generated by the processing evaluation unit 12 in association with the processing condition corresponding to the processing condition, and furthermore, the processing condition corresponding to the prediction value generated by the prediction unit 27 . are displayed in correspondence with each other, and the optimum processing conditions are indicated.

Next, the hardware configuration of the processing condition search device 2 will be described. The search processing condition generation unit 11, the processing evaluation unit 12, the machine learning unit 13, and the optimum processing condition generation unit 14 shown in Fig. 1 are each realized by a processing circuit. The processing circuit may be dedicated hardware or a circuit including a processor. The display unit 15 is realized by display means such as a display or monitor, a processing circuit for displaying display data on the display means, and input means such as a keyboard and a mouse.

When the processing circuit is a circuit including a processor, the processing circuit is, for example, a processing circuit of the configuration shown in FIG. 2 . 2 is a diagram showing a configuration example of a processing circuit according to the present embodiment. The processing circuit 200 includes a processor 201 and a memory 202 . When each part of the processing condition search apparatus 2 is realized by the processing circuit 200 shown in FIG. 2, the processor 201 reads and executes the program stored in the memory 202, and thus the search end determination unit 21 , processing condition generation unit 22 , actual processing command unit 23 , processing result acquisition unit 24 , evaluation value generation unit 25 , prediction unit 27 , uncertainty evaluation unit 29 and optimum The function of the processing condition calculation unit 31 is realized. That is, when each part of the processing condition search device 2 is realized by the processing circuit 200 shown in Fig. 2, these functions are realized using a program that is software. The search result storage unit 26 , the prediction result storage unit 28 , the uncertainty storage unit 30 , and the optimum processing condition storage unit 32 are realized by the memory 202 . The memory 202 is also used as a working area of the processor 201 . The processor 201 is a CPU (Central Processing Unit) or the like. The memory 202 corresponds to, for example, a non-volatile or volatile semiconductor memory such as a random access memory (RAM), a read only memory (ROM), a flash memory, or a magnetic disk.

When the processing circuit realizing the search processing condition generation unit 11, the processing evaluation unit 12, the machine learning unit 13 and the optimum processing condition generation unit 14 of the processing condition search device 2 is dedicated hardware, The processing circuit is, for example, an FPGA (Field Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit). In addition, each part of the processing condition search apparatus 2 may be implemented by combining the processing circuit provided with a processor, and dedicated hardware. Each unit of the processing condition search device 2 may be realized by a plurality of processing circuits.

Next, the operation of the present embodiment will be described. 3 is a flowchart showing an example of a processing condition search processing procedure in the processing condition search apparatus 2 according to the present embodiment. When the machining condition search processing is started, first, the machining condition generating unit 22 generates initial machining conditions (step S1). The machining condition generating unit 22 generates initial machining conditions by selecting a predetermined number of machining conditions as initial machining conditions from all combinations that can be set as machining conditions. As a method of selecting initial processing conditions, any method may be used, and examples of design of experiments, optimal programming, random sampling, and the like are exemplified. In addition, when the user can guess the processing conditions considered to be optimal, such as past usage records, you may use the processing conditions input by the user as initial processing conditions.

For example, if there are five control parameters constituting the processing conditions, and a value to be set in the processing machine 1 can be selected from the values of 11 steps for each control parameter, the total number of combinations of processing conditions is 16,1051. Among these combinations, ten processing conditions are selected as initial processing conditions. In addition, the number of control parameters constituting the processing conditions, the number of steps that can be set for each control parameter, and the number of processing conditions to be selected as the initial processing conditions are not limited to these. The number of steps that can be set may differ depending on the control parameter.

Next, the processing condition search device 2 selects one initial processing condition from the generated initial processing conditions, and causes the processing machine 1 to process under the selected initial processing conditions (step S2). In detail, the machining condition generating unit 22 selects one of the initial machining conditions, and outputs the selected initial machining condition to the actual machining command unit 23 and the evaluation value generating unit 25 . The real processing command unit 23 generates a command for operating the processing machine 1 based on the initial processing conditions input from the processing condition generation unit 22 , and outputs the generated command to the processing machine 1 . As a result, the processing machine 1 performs processing based on the initial processing conditions selected by the processing condition generation unit 22 . As described above, the processing condition search apparatus 2 of the present embodiment first causes the processing machine 1 to perform processing based on the initial processing conditions. Processing based on initial processing conditions is hereinafter also referred to as initial processing.

Next, the machining condition search device 2 generates an evaluation value corresponding to the machining performed in step S2 (step S3). In detail, the processing result acquisition unit 24 acquires information indicating a processing result corresponding to the processing performed in step S2 and outputs it to the evaluation value generation unit 25 . The evaluation value generation unit 25 obtains an evaluation value indicating the quality of the processing performed in step S2 by using the information indicating the processing result.

The evaluation value is a value from 0 to 1, for example, as described above. A continuous value may be sufficient as an evaluation value, and the discrete value called 10-step evaluation may be sufficient as it. For example, the evaluation value generation unit 25 calculates and calculates the amount of processing defects such as dross, oxide film peeling, roughness, and scratches by image processing the image of the cut surface of the work. An evaluation value indicating the good or bad state of processing is calculated according to the quantity. Any method may be used for calculating the amount of dross, oxide film peeling, roughness, and scratches. For example, the evaluation value generating unit 25 calculates the area in which these occur due to the difference in brightness of pixels, , the amount of machining defects occurring according to the length or area of the region is quantitatively calculated. And the correspondence between the evaluation value and the quantity in which the processing defect is occurring is determined so that the evaluation value becomes low as there is more quantity in which the processing defect is produced, and the evaluation value generation part 25 calculates an evaluation value based on this correspondence . As described above, the evaluation value may be input by the user. Moreover, an evaluation value may be determined according to the sound and light which generate|occur|produced during processing. For example, you may define an evaluation value so that an evaluation value may become small, so that the level of the sound generated during processing is large. Moreover, the evaluation value generation|generation part 25 may combine several types of information and may calculate an evaluation value.

Next, the machining condition search device 2 stores machining conditions and evaluation values (step S4). Specifically, the processing result acquisition unit 24 associates the processing condition received from the processing condition generation unit 22 with the evaluation value calculated in step S3, and stores it in the search result storage unit 26 as a search result. .

Next, the processing condition search device 2 determines whether or not to end the initial processing (step S5), and when the initial processing is finished (step S5 YES), the processing of steps S6 and S8 to be described later is performed. . In step S5, specifically, the machining condition search device 2 determines whether or not machining corresponding to all the initial machining conditions generated in step S1 has been performed. The machining condition search device 2 determines that the initial machining is finished when machining corresponding to all the initial machining conditions generated in step S1 is performed.

In the case where the initial processing is not finished (No in step S5), the processing condition search device 2 repeats the processing from step S2. In the second and subsequent steps S2, the machining condition search device 2 selects the initial machining conditions not selected in the previous step S2. In addition, here, the example in which the processing condition search apparatus 2 made the processing machine 1 perform processing corresponding to one initial processing condition, and calculated|required and memorize|stored the evaluation value was demonstrated. Not limited to this example, the processing condition search device 2 causes the processing machine 1 to perform processing corresponding to all the generated initial processing conditions, and transmits information indicating processing results corresponding to such processing to the processing machine 1 ), an evaluation value for each initial processing condition may be obtained and stored.

In step S6, the prediction unit 27 uses the search result stored in the search result storage unit 26 to calculate a predicted value of the evaluation value corresponding to the unsearched processing condition, which is a processing condition that has not yet been processed. . Regarding the unsearched processing condition, which is the processing condition in which processing has been performed, an evaluation value is generated by the above-described step S3. On the other hand, the processing conditions under which processing is performed are a part of a combination of all processing conditions. For example, if all combinations of processing conditions are 161041 types, and 10 types of initial processing conditions are generated, after the end of the initial processing, there are 161031 types of unexplored processing conditions. Accordingly, in this case, the prediction unit 27 calculates 16,1031 predicted values of the evaluation values. Moreover, as mentioned later, also in step S13 - step S15, selection of a process condition, execution of a process, and calculation of an evaluation value are performed, and step S6 is implemented after step S15. When step S6 is executed via steps S13 to S15, the processing conditions set in step S13 are excluded from the unsearched processing conditions.

As an example of the method in which the prediction unit 27 calculates the predicted value of the evaluation value corresponding to the non-explored processing condition, that is, the method of predicting the evaluation value corresponding to the non-explored processing condition, a method using Gaussian process regression is exemplified. When the prediction unit 27 predicts the evaluation value corresponding to the unexplored processing condition using Gaussian process regression, the following calculation is performed. The method using Gaussian process regression is an example of a method of using a probabilistic model for the processing condition of the evaluation value generated assuming that the evaluation value for the processing condition is a random variable according to a specific distribution. Let N be the number of observation values, that is, the number of processing conditions under which processing is performed and evaluation values are calculated, let the gram matrix be C _N , and let t be a vector listing the evaluation values recorded in the search result storage unit 26 . . At this time, one of the control parameters constituting the machining condition is set to x _i (i is a natural number), and the value of this control parameter for each machining condition recorded in the search result storage unit 26 is x ₁ to x _{When N} is set, the predicted value m( _xN+1 ) of the evaluation value with respect to the processing condition _xN+1 of unexplored can be calculated by the following formula (1). k is, as shown in the following formula (2), the searched processing condition x ₁ , ... each of x and _N is the vector lists the value of the kernel function of when to acquire x _{N + 1.} Also, the superscript T denotes a transposition, and the superscript -1 denotes an inverse matrix.

m(x _N+1 )=k ^{T .} (C _N ^-1 ) t ... (1)

In addition, here, although the prediction part 27 demonstrated the example in which prediction using Gaussian process regression was demonstrated, the prediction method of an evaluation value is not limited to this, For example, a decision tree, linear regression, boosting, and a neural A method using machine learning with a teacher called a neural network may be used.

Next, the processing condition search device 2 records the predicted value of the evaluation value (step S7). Specifically, the prediction unit 27 associates the predicted value of the evaluation value calculated in step S6 with the processing conditions, and stores it in the prediction result storage unit 28 as a prediction result.

Further, the uncertainty evaluation unit 29 uses the search result stored in the search result storage unit 26 to calculate an index indicating the uncertainty with respect to the prediction of the evaluation value corresponding to the unsearched processing condition (step S8). . As an example of an index indicating uncertainty, a standard deviation calculated using Gaussian process regression, which is an example of a probabilistic model, may be cited. When the uncertainty evaluation unit 29 calculates an index indicating uncertainty using Gaussian process regression, for example, the following calculation is performed. Let N be the number of observation values, i.e., the number of processing conditions for which processing is performed and evaluation values are calculated, let the gram matrix be C _N , and let k be a vector listing the processing conditions recorded in the search result storage unit 26, , let c be a scalar value obtained by adding the precision parameter of the prediction model to the kernel value of unexplored processing conditions x _N+1. At this time, one of the control parameters constituting the machining condition is set to x _i (i is a natural number), and the value of this control parameter for each machining condition recorded in the search result storage unit 26 is x ₁ to x _{When N is set, the standard deviation σ(x N+1} ), which is an index indicating the uncertainty in the prediction of the evaluation value for the unexplored processing condition x _N+1 , can be calculated by the following formula (3). Moreover, although the variance sigma ² (x _{N+1 ) is calculated|required in Formula (3), the standard deviation (sigma)(x N+1} ) can be calculated|required by calculating the square root of a variance.

σ ² (x _N+1 )=c-k ^T ·(C _N ^-1 )·k ··· (3)

Incidentally, here, an example has been described in which the uncertainty evaluation unit 29 calculates an index indicating uncertainty about prediction using Gaussian process regression, but the method of calculating the index indicating uncertainty is not limited to this, and examples For example, a method such as density estimation or a mixed density network may be used.

Here, the predicted value of the evaluation value and the uncertainty of the predicted value in this embodiment are demonstrated. 4 is an explanatory diagram conceptually indicating a predicted value of an evaluation value of the present embodiment and an index indicating uncertainty. 4 shows an example in which a predictive value and an index indicating uncertainty are calculated using Gaussian process regression. The horizontal axis of FIG. 4 represents the value x of the control parameter that is the processing condition, and the vertical axis of FIG. 4 represents the evaluation value. Points indicated by black dots in FIG. 4 indicate evaluation values (hereinafter also referred to as evaluation values of thread processing) calculated based on thread processing using initial processing conditions. In the prediction using Gaussian process regression, the evaluation value is predicted, assuming that the evaluation value follows a Gaussian distribution. For this reason, if the predicted value of the evaluation value is the average m(x) of the Gaussian distribution, and the index indicating the uncertainty of the prediction is the standard deviation σ(x) of the Gaussian distribution, the actual evaluation value is approximately 95% , m(x)-2σ(x) or more and m(x)+2σ(x) or less fall within the range statistically. In Fig. 4, the curve indicated by the solid line represents m(x), which is the predicted value of the evaluation value, and the curve indicated by the dotted line is the curve of m(x)-2σ(x) and m(x)+2σ(x). indicates. As shown in FIG. 4 , the index indicating uncertainty becomes small at a location close to the evaluation value of real processing, and the index indicating uncertainty increases at a location distant from the evaluation value of real processing.

Returning to the description of FIG. 3 , after step S8 , the processing condition search device 2 records an index indicating the uncertainty of the predicted value (step S9 ). In detail, the uncertainty evaluation unit 29 stores the index indicating the uncertainty of the predicted value calculated in step S8 in correspondence with the processing conditions in the uncertainty storage unit 30 .

After steps S7 and S9, the machining condition search device 2 calculates the optimum machining conditions (step S10). Specifically, the optimum processing condition calculation unit 31 uses the search result stored in the search result storage unit 26 and the prediction result stored in the prediction result storage unit 28 so that the evaluation value is equal to or greater than the threshold value. , and also calculate the processing condition with the highest tolerance, that is, the highest robustness. The threshold value can be set to 1, which is a value indicating that processing of the highest quality has been performed when the evaluation value takes a value from 0 to 1. Alternatively, the threshold value may be set to a value close to 1 and less than 1, such as 0.85.

Here, in the present embodiment, the tolerance means that even if there is a deviation in the evaluation value obtained by processing by the processing machine 1 using the same processing conditions, when a certain processing condition is selected and processing is performed, the selected It represents the height of the possibility that an evaluation value equivalent to the evaluation value corresponding to processing conditions will be obtained. The evaluation value corresponding to the selected processing condition is an evaluation value obtained by real processing and a predicted value of the evaluation value, and these evaluation values are hereinafter also referred to as assumed evaluation values. In this embodiment, although processing conditions are selected based on the estimated evaluation value so that it may mention later, it is selected so that the estimated evaluation value selected at this time may become more than a desired value. On the other hand, the evaluation value obtained by actually processing based on the processing conditions selected in this way may differ from the estimated evaluation value by some factors. In this embodiment, the height of the possibility that the evaluation value obtained by actually processing is performed based on the estimated evaluation value and the selected processing conditions matches as a tolerance is defined as a tolerance, By selecting a processing condition with a high tolerance, the same Even if there exists a dispersion|variation in the evaluation value obtained by the processing machine 1 processing using processing conditions, the possibility that the processing used as a desired evaluation value will be performed is raised. As an example of the allowable error, in a multidimensional space in which the parameter values of a plurality of control parameters constituting the processing conditions are the dimensions, a point corresponding to an arbitrary processing condition in which the evaluation value is equal to or greater than the threshold value, the evaluation value is equal to or greater than the threshold value A value indicating how far away from the boundary between the region where , and the region where the evaluation value is less than the threshold value is given. In this case, for example, the processing condition used as the center of the area|region where the evaluation value becomes more than a threshold value can be defined as the processing condition with the highest allowable error.

5 : is a figure for demonstrating the allowable error of this embodiment. 5 shows a plane in which the parameter values of each of the first control parameter and the second control parameter are taken on the abscissa axis and the ordinate axis when the processing condition is composed of two control parameters of the first control parameter and the second control parameter, have. In Fig. 5, points indicated by circles, triangles, and squares indicate points corresponding to the processing conditions under which real processing was performed. Processing conditions 301 indicated by circles indicate processing conditions in which the evaluation value is 0.85 or more. Processing conditions 302 indicated by triangles indicate processing conditions in which the evaluation value is 0.45 or more and less than 0.85. Processing conditions 303 indicated by squares indicate processing conditions in which the evaluation value is less than 0.45. This evaluation value is an evaluation value calculated by real processing when it corresponds to the point where actual processing was performed, and is the above-mentioned predicted value in a point corresponding to the unexploited processing condition.

Here, for example, the case where the evaluation value is 0.85 or more is called good processing, the case where the evaluation value is 0.45 or more and less than 0.85 is called mild bad processing, and the case where the evaluation value is less than 0.45 is called severe bad processing. In FIG. 5 , a region 311 is a region in which an evaluation value is 0.85 or more, a region 312 is a region in which an evaluation value is 0.45 or more and less than 0.85, and a region 313 is a region in which an evaluation value is 0.45. area that is less than When the above-described threshold value is 0.85 and it is desirable to set processing conditions such as good processing, the boundary between the region in which the evaluation value is equal to or greater than the threshold value and the region in which the evaluation value is less than the threshold value is region 311 and a boundary of the region 312 . In this case, the farther away from the boundary between the region 311 and the region 312, the higher the allowable error. Therefore, for example, the center 320, which is the center of gravity of the region 311, becomes the processing condition with the highest allowable error. In this case, the optimum processing condition calculation unit 31 optimizes the processing conditions corresponding to the center of gravity of the region in which the evaluation value is equal to or greater than the threshold value in a multidimensional space in which a plurality of control parameters constituting the processing conditions are dimensions. It is obtained as a processing condition.

The calculation method of the processing condition with the highest allowable error is not limited to the example mentioned above. For example, weighting is given according to the value of the evaluation value, and the processing condition corresponding to the point serving as the coordinate obtained by taking the weighted average may be regarded as the processing condition with the highest allowable error. For example, In the example shown in Fig. 5, the weight of not less than a threshold value above 0.45, and the evaluation value is 0.85 to w _1, and the evaluation value is 0.45 or more and, and the weight for the case of 0.85 lower than in w ₂ , w ₁ + w ₂ =1. In this case, if the coordinates of the center of gravity of the region 311 are p ₁ , and the coordinates of the center of gravity of the region 312 are p ₂ , then the coordinate p _L with the highest allowable error is expressed in Equation (4) below. can be obtained by

p _L =w ₁ ×p ₁ +w ₂ ×p ₂ ... (4)

In addition, the calculation method of a weighted average is not limited to the above-mentioned example. Further, as described above, the predicted values are calculated for all unsearched processing conditions, and the above-described area is calculated using these results. Therefore, in implementation, the predicted value of the evaluation value is computed with respect to the discretely arrange|positioned point, rather than a continuous boundary computed as shown in FIG. Accordingly, for example, when the coordinates of the center of gravity of the area 311 are calculated, the center of gravity of each coordinate value of a point discretely arranged in the area 311 is calculated. The same is true for weighted averages. For this reason, when the central point, which is the center of gravity, or the weighted average is obtained, the amount of calculation increases when the number of combinations of processing conditions is large. Therefore, in order to reduce the amount of calculation, the above-described central point and weighted average may be calculated after the processing conditions to be considered are thinned out.

In the present embodiment, as described above, since, for example, the center 320, which is the center of gravity of the region 311, is selected as the optimum processing condition with the highest tolerance, the optimum processing condition is the area 311 ) away from the boundary. On the other hand, as in Patent Document 1, when the point at which the evaluation value becomes the highest is obtained by searching for real machining, and searching by real machining is performed centering on that point, the area 330 is searched, for example. . The region 330 is a region skewed to the upper right of the paper in the region 311, and even if an optimal processing condition is searched for in this region, the region 311 is higher than the optimal processing condition selected in the present embodiment. Only machining conditions close to the boundary can be searched. As described above, in the present embodiment, since the processing condition remote from the boundary of the region 311 can be selected as the optimum processing condition, the desired processing result can be obtained even when there is a variation in the processing result obtained by the same processing condition. increase the likelihood that

Returning to the description of FIG. 3 , after step S10 , the machining condition search device 2 stores the optimum machining conditions (step S11 ). Specifically, the optimum machining condition calculation unit 31 stores the optimum machining condition obtained in step S10 in the optimum machining condition storage unit 32 .

Next, the processing condition search device 2 determines whether or not to end the search (step S12), and when the search ends (YES in step S12), the processing ends. Specifically, in step S12, the search end determination unit 21 determines whether the search end condition is satisfied. Examples of the search termination condition include a condition that, for all processing conditions, an index indicating the uncertainty of the prediction of the evaluation value is equal to or less than a threshold value. Further, the search end determination unit 21 may calculate a prediction error rate, which is a probability that the predicted value is incorrect, based on the index indicating uncertainty and the predicted value, and determine whether or not to end the search for processing conditions based on the prediction error rate. . In addition, as an example of the search termination condition, a condition that the number of processing conditions that may significantly deviate from the prediction is less than or equal to the specified number is exemplified. This is because the predicted value is calculated based on a plurality of actual measured values in the vicinity, so if there is a processing condition in which the actual value and the predicted value are significantly different, the actual value of the processing condition affects a wide range of the search space, and the predicted value changes significantly. to be.

For example, let the evaluation value be a value from 0 to 1, and it is assumed that 1 is a value indicating that the best processing was performed. In the range of evaluation values from 0 to 1, it is undesirable that the evaluation value predicted by the user and the evaluation value obtained in actual processing differ greatly. For example, assuming that machining with an evaluation value of 0.85 or higher is judged as good machining, actual machining is performed under the machining conditions predicted by the evaluation value of 0.95, and even if the evaluation value obtained by this machining is 0.9, the effect on product quality is less On the other hand, if actual machining was performed under the machining conditions predicted by the evaluation value of 0.95, and the evaluation value obtained by this machining was 0.2, the machining would have been performed under the machining conditions of good quality, but this would result in severe machining defects and have a large effect. In this way, when there are two ranges that have a large effect when the range to which the evaluation value obtained from actual machining and the range to which the predicted value belongs, such as the range of severe machining failure and the range judged to be good machining, there are two ranges. The prediction error rate, which is the probability of occurrence, may be used to determine the end of the search.

For example, the processing conditions under which the evaluation value falls within the first range are group #1, and the machining conditions under which the evaluation value falls within the second range are group #1. A 1st range is 0 or more and less than 0.3, for example, A 2nd range is a range which does not overlap with a 1st range, For example, it is 0.7 or more and 1 or less. In this case, when processing is performed under the processing conditions predicted to belong to group #1, it is not preferable to actually become an evaluation value corresponding to group #1. Moreover, when processing is carried out under the processing conditions predicted to belong to group #1, it is not preferable to actually become an evaluation value corresponding to group #1. Accordingly, the search end determination unit 21 uses the prediction result stored in the prediction result storage unit 28 and the index indicating the uncertainty of the prediction stored in the uncertainty storage unit 30, the first prediction error rate and the second prediction error rate. Calculate the prediction error rate. The first prediction error rate is the probability that the machining conditions predicted to belong to group #1 actually belong to group #1, and the second prediction error rate is the probability that the machining conditions predicted to belong to group #1 actually belong to group #1. probability of belonging to In other words, the first prediction error rate is a probability that the actual evaluation value of the processing condition predicted that the predicted value is within the first range falls within the second range that does not overlap the first range. Here, it is assumed that Gaussian process regression is used, and it is assumed that the predicted value m(x) of the evaluation value for the arbitrary processing condition x is 0≤m(x)≤0.3, that is, the arbitrary processing condition x belongs to group #1. Let the standard deviation of the uncertainty about the prediction be σ(x). At this time, the first prediction error rate, which is the probability that the processing condition x actually belongs to group #1 instead of group #1, can be calculated by the following equation (5). The second prediction error rate can likewise be calculated by changing the integration range. z is a variable for performing integration.

The search end determination unit 21 may determine that the search is complete when both the first prediction error rate and the second prediction error rate are equal to or less than a predetermined value, or may determine that the search is complete when the first prediction error rate is less than or equal to a predetermined value. The first prediction error rate and the second prediction error rate calculated in this way are examples of the prediction error rate indicating the probability that the prediction is erroneous. The predicted erroneous interest rate may be an index relating to the difference between prediction and actual measurement, and is not limited to the above-described example. For example, the prediction error rate may be an index indicating the uncertainty of prediction of the above-described evaluation value. In addition, the search end determination unit 21 combines a plurality of items, such as determining that the search has been completed, when the index indicating the uncertainty of prediction of the evaluation value is equal to or greater than the threshold and the first prediction error rate is equal to or less than a predetermined value. You may decide to end the search.

If it is determined in step S12 that the search is not to be ended (NO in step S12), the machining condition search device 2 generates machining conditions to be searched for next (step S13). In detail, when the search end determination unit 21 determines that the search end condition is not satisfied, it instructs the machining condition generation unit 22 to generate a machining condition to be searched next, and the machining condition generation unit 22 ) selects a machining condition to be searched next, that is, a new machining condition, from among all machining conditions, and outputs the selected machining condition to the actual machining command unit 23 . The search end determination unit 21 determines whether or not to end the search for processing conditions, for example, based on an index indicating uncertainty, and, when not ending the search, sends a new Create processing conditions.

As an example of the selection method when the search end determination unit 21 selects the processing condition to be searched for next, a method of selecting a processing condition corresponding to the evaluation value for the processing result at the boundary between good processing and poor processing is exemplified. can When the region in the multidimensional space of the processing conditions corresponding to the evaluation value for the processing result at the boundary between good processing and poor processing becomes clear, the inner side of the region is defined as a region with a high evaluation value, and the outer side becomes a region with a low evaluation value. It can be determined, and the central coordinates of the region where the evaluation value increases with a small number of trials can be obtained.

For example, when a processing condition in which the evaluation value is around 0.7 is selected, the predicted value of the evaluation value for the processing condition x is m(x), and the standard deviation, which is an index indicating uncertainty about the prediction, is σ(x). . At this time, if the search end determination unit 21 selects x for which the value of the following equation (6) is the maximum, processing conditions with an evaluation value close to 0.7 can be obtained.

-(m(x)-0.7) ² +κ σ(x) ... (6)

Here, κ is a parameter that determines the weighting of the prediction and its uncertainty, and as κ decreases, the search becomes more focused on prediction, and as κ increases, the search is focused on uncertainty. The target evaluation value may be fixed to the same value from the start to the end of the search, or may be set to a different value each time the search is performed. In addition, when the search end determination unit 21 calculates the above-described prediction error rate, the prediction error rate may be obtained from the search end determination unit 21 and processing conditions may be selected within a range where the prediction error rate is high. Alternatively, the processing condition generating unit 22, similarly to the search end determination unit 21, is assigned to the prediction result stored in the prediction result storage unit 28 and the index indicating the uncertainty of the prediction stored in the uncertainty storage unit 30. A prediction error rate may be calculated|required based on the prediction error rate, and a machining condition may be created based on the calculated|required prediction error rate. For example, the machining condition generation unit 22 selects machining conditions from within a range with a high prediction error rate.

After step S13, similarly to steps S2 and S3, steps S14 and S15 are executed, and after step S15, steps S6 and S8 are executed again.

The display unit 15 can display data obtained in the process of the above-described processing, optimal processing conditions obtained as a result of the processing, and the like. For example, the display unit 15 associates and displays the evaluation value generated by the processing evaluation unit 12 and the corresponding processing condition, and furthermore, the processing condition corresponding to the predicted value generated by the prediction unit 27 . can be displayed in correspondence with each other, and the optimum processing conditions can also be displayed. 6 : is a figure which shows the display example of the processing condition and evaluation value of this embodiment. In the example shown in FIG. 6, each of two control parameters is taken on a horizontal axis and a vertical axis|shaft, and the evaluation value in each point which shows a processing condition is shown with the difference of a figure. In Fig. 6 , points corresponding to the hatched graphic indicate processing conditions that have been searched for, and points corresponding to the non-hatched graphic indicate predicted processing conditions. In addition, a circle figure indicates that the evaluation value is an evaluation value corresponding to good processing, a figure of a square indicates that the evaluation value is an evaluation value corresponding to poor processing, and a triangular figure indicates that the evaluation value is an evaluation value corresponding to poor processing. It shows that it is an evaluation value to do. Dots with double circles indicate the estimated optimal processing conditions. In addition, in the background of the point which shows each processing condition in FIG. 6, the range of favorable processing is shown with vertical hatching. The display example of processing conditions and evaluation value is not limited to the example shown in FIG.

In the example demonstrated above, although the processing machine 1 demonstrated it as a laser processing machine, as mentioned above, the processing machine 1 is not limited to a laser processing machine. When the search method of the processing conditions demonstrated in this embodiment is applied to the processing machine 1 other than a laser processing machine, the processing conditions comprised by the control parameter according to the processing machine 1 can be searched similarly. For example, when the machining machine 1 is a wire electric discharge machine, a discharge pause time, a discharge pulse energy, an average machining voltage, etc. can be made into control parameters.

As described above, according to this embodiment, the machining condition search device 2 includes a machining condition generating unit 22 that generates machining conditions composed of one or more control parameters settable in the machining machine 1 ; A processing evaluation unit 12 is provided that calculates an evaluation value indicating the quality or failure of the performed processing, based on the information for evaluating the processing performed based on the processing conditions generated by the condition generation unit 22 do. Further, the processing condition search device 2 includes a prediction unit 27 that predicts the evaluation value of the processing condition in which the corresponding processing is not performed, based on the evaluation value calculated by the processing evaluation unit 12 . . Further, the processing condition search device 2 uses the prediction value predicted by the prediction unit 27 to calculate the optimum processing condition calculation unit 31 for calculating the processing condition in which the evaluation value is equal to or greater than the threshold value and the tolerance is high. be prepared Thereby, with a small number of trials, it is possible to obtain a processing condition with a high possibility of obtaining a processing result equal to or higher than a desired evaluation value.

Further, the processing condition search device 2 includes an uncertainty evaluation unit 29 that calculates an index indicating the uncertainty of prediction by the prediction unit 27, and uses this index to determine whether the search termination condition is satisfied. When the determination is made and the search is not terminated, processing conditions for performing actual processing are newly created, and an evaluation value corresponding to the processing performed is calculated based on the generated processing conditions. Then, until the search end condition is satisfied, the processing condition generation unit 22 , the actual processing command unit 23 , the prediction unit 27 , the uncertainty evaluation unit 29 , and the optimum processing condition calculation unit 31 process Repeat. Accordingly, it is possible to reduce the number of processing conditions in which thread processing is performed as an initial search, and optimal processing conditions can be obtained with the minimum number of times of thread processing that satisfies the search termination condition.

The configuration shown in the above embodiment shows an example of the content of the present invention, it can be combined with other known techniques, and without departing from the gist of the present invention, it is also possible to omit or change a part of the configuration. It is possible.

1: Machining machine 2: Machining condition search device
11: search processing condition generation unit 12: processing evaluation unit
13: machine learning unit 14: optimal processing condition generation unit
15: display unit 21: search end determination unit
22: processing condition generation unit 23: thread processing command unit
24: processing result acquisition unit 25: evaluation value generation unit
26: search result storage unit 27: prediction unit
28: prediction result storage unit 29: uncertainty evaluation unit
30: uncertainty storage unit 31: optimum processing condition calculation unit
32: optimum processing condition storage unit

Claims (10)

A machining condition generating unit that generates machining conditions consisting of one or more control parameters that can be set on the machine;
a thread processing command unit for causing the processing machine to perform processing based on the processing conditions generated by the processing condition generation unit;
a processing evaluation unit configured to generate an evaluation value of the performed processing based on information indicating a processing result of the processing performed;
a prediction unit for predicting an evaluation value corresponding to a processing condition in which processing is not performed, based on the evaluation value and the processing condition corresponding to the evaluation value;
Based on the predicted value predicted by the prediction unit and the evaluation value generated by the processing evaluation unit, calculating the optimum processing condition for obtaining the optimum processing condition, which is the processing condition in which the evaluation value is equal to or greater than the threshold value and the allowable error is maximized wealth
to provide
The evaluation value, characterized in that the larger the value, the better the processing result
Machining condition search device. The method of claim 1,
an uncertainty evaluation unit for calculating an index indicating uncertainty of prediction by the prediction unit;
Based on the index, it is determined whether or not to end the search for processing conditions, and when the search is not ended, a search end determination unit that causes the processing condition generation unit to generate a new processing condition.
to provide
The processing of the processing condition generating unit, the actual processing instruction unit, the processing evaluation unit, the prediction unit, the uncertainty evaluation unit, and the optimum processing condition calculation unit is performed until it is determined by the search end determination unit that the search is finished. characterized by repeating
Machining condition search device. 3. The method of claim 2,
The search end determination unit calculates a prediction error rate, which is a probability that the predicted value is incorrect, based on the index and the predicted value, and determines whether to end the search for processing conditions based on the prediction error rate. navigation device. 4. The method of claim 3,
The search end determination unit is a first prediction that is a probability that, based on the index and the predicted value, the actual evaluation value of the processing condition predicted that the predicted value is within the first range falls within a second range that does not overlap with the first range A machining condition search apparatus, comprising: calculating an error rate, and determining whether to end the search for machining conditions based on the first prediction error rate. 5. The method according to any one of claims 2 to 4,
The machining condition generating unit generates a machining condition in which an evaluation value of the machining condition becomes an intermediate value, based on the predicted value. 6. The method according to any one of claims 2 to 5,
The machining condition generating unit, machining condition search apparatus, characterized in that by using the prediction error rate, which is a probability that the predicted value is wrong, to generate the machining conditions. 7. The method according to any one of claims 2 to 6,
The prediction unit generates the prediction value using a probabilistic model for the processing condition of the evaluation value, which is generated assuming that the evaluation value for the processing condition is a random variable according to a specific distribution,
wherein the uncertainty evaluation unit calculates the index by using the probabilistic model
Machining condition search device. 8. The method according to any one of claims 1 to 7,
The optimum processing condition calculation unit is configured to obtain, as the optimum processing condition, a processing condition corresponding to a center of gravity of a region in which the evaluation value is equal to or greater than a threshold in a multidimensional space in which a plurality of control parameters constituting the processing conditions are dimensions. A processing condition search device, characterized in that. 9. The method according to any one of claims 1 to 8,
A display unit for displaying the evaluation value generated by the processing evaluation unit in correspondence with the processing condition corresponding to it, and displaying the prediction value generated by the prediction unit in correspondence with the processing condition corresponding to the processing condition, and displaying the optimum processing condition A processing condition search device, characterized in that it comprises a. A processing condition search device,
A machining condition creation step of creating a machining condition consisting of one or more control parameters that can be set in the machine;
a thread processing instruction step of causing the processing machine to perform processing based on the processing conditions generated by the processing condition generation step;
a processing evaluation step of generating an evaluation value of the performed processing based on information indicating a processing result of the processing performed;
a prediction step of predicting an evaluation value corresponding to a processing condition in which processing is not performed, based on the evaluation value and the processing condition corresponding to the evaluation value;
Based on the predicted value predicted by the predicting step and the evaluation value generated by the machining evaluation step, calculating the optimum machining condition to obtain the optimum machining condition, which is the machining condition in which the evaluation value is equal to or greater than the threshold value and the allowable error is maximized step
including,
The evaluation value, characterized in that the larger the value, the better the processing result
How to search for machining conditions.

Images